Abstract
Purpose: :
To evaluate an automated computer algorithm for the segmentation of the optic canal in 3D OCT images. Segmentation of the optic canal has the potential to provide a continuous, stable reference plane for the measurement of structural parameters in identifying progression in early glaucomatous damage.
Methods: :
30 patients with glaucoma underwent SD-OCT ONH imaging (60 volumes, 1 per eye, 200x200x1024 voxels) using a CirrusTM HD-OCT machine (Carl Zeiss Meditec, Inc., Dublin, CA). Our graph search based multilayer segmentation algorithm was used to identify individual layers surrounding the optic canal reference plane. A 2D projection image was obtained by averaging the voxel values between two segmented surfaces parallel to the reference plane and the vessels of the projection image were segmented automatically. The vessel pixels were identified and removed in the projection image by interpolation of surrounding non-vessel pixels. An optimal graph search algorithm was applied to segment the optic canal boundary from this ‘vessel-suppressed’ projection image. A reference standard was obtained by combining three expert manual segmentations on corresponding stereo fundus disc photographs and registering them with the OCT images.
Results: :
The segmentation result was compared with the reference standard. We found an average unsigned minimum distance of 0.137± 0.061 mm and an average signed minimum distance of -0.027±0.077 mm for the 60 scans (with the negative sign meaning our segmentation was slightly closer, on average, to the center of the ONH).
Conclusions: :
A novel automated vessel suppression based segmentation algorithm can segment the optic canal on 3D OCT images with a high degree of accuracy.
Keywords: image processing • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • imaging/image analysis: clinical